Upload files

how-to · Reviewed 2026-07-31 · FEAT-0117 GA

Three ways to get data in. The browser suits small sets; above a few gigabytes, use the Transfer Manager or the CLI.

Pick a method

Method	Use it for	Avoid when
Browser	A handful of files, each under a few GB	Large files, many files, flaky networks
Rescale CLI	Scripted, resumable, multi-gigabyte uploads	—
Transfer Manager	Large datasets from a Windows desktop	You're on macOS or Linux (use the CLI)

The dividing line is roughly 80 MB per file: below it the simple browser path is fastest, above it chunked transfer wins. The platform switches automatically in the browser, but for large sets the CLI or Transfer Manager is more reliable because interrupted transfers resume instead of restarting.

Upload through the browser

On the Files tab, or while attaching inputs during job setup, click the upload control and select files or a folder.

To upload a whole directory, select the folder rather than its contents. The browser preserves the structure, so nested directories stay intact.

Upload with the CLI

rescale-cli upload -p -f <path>

The CLI handles chunking, retries, and parallelism. Use it for anything scripted or over a few gigabytes.

To upload straight to CFS instead of Rescale Files, add --copy-to-cfs.

When a file becomes a job input

This trips people up during job setup:

• Uploading a file on the job-setup page saves it as an input file immediately.
• It isn't associated with the job until you save the job.

So an abandoned job setup can leave uploaded files in your My Library, they're stored and billed even though no job uses them. Check your library after abandoning a setup.

Attach inputs after adding software

If you add input files before configuring any software tile, they attach to the first tile you add afterwards. On multi-software jobs that may not be where you want them.

Compressed files are unpacked automatically

Upload an archive as a job input and Rescale extracts it when the job starts. Supported: .tar.gz, .tar.bz2, .tar, .gz, .bz2, .tgz, .zip, .7z, .7zip.

Compressing before upload is worth doing. It cuts both transfer time and file count, and matters most above about 10 GB.

Filename and path limits

Limit	Value
Name plus path	200 bytes — 200 ASCII characters, or about 100 in Japanese, Korean, or other multi-byte text
API single-file upload	5 GB

Exceeding the path limit loses data

A file whose name and path exceed 200 bytes cannot be saved. On job output this means results are silently lost, so keep generated paths short in deeply nested multi-run jobs.

Windows users hitting the separate 260-character path limit on download should see Transfer Manager troubleshooting.

Troubleshooting

• Stalls or fails silently in the browser — try another browser, disable upload-blocking extensions, check your corporate proxy. Troubleshooting browser file upload issues
• Many files or folders at once — the UI accepts drag-and-drop, but the CLI is faster and more reliable at scale. Uploading folders and multiple files
• Blocked by a limit — a refused upload may be an enforced cap, not a bug. See limits and policies.

Related

• Download files
• Transfer Manager — bulk transfer on Windows
• Browse your files — where uploads land
• Limits and policies